[GitHub] [incubator-tvm] nolanliou edited a comment on issue #6268: TVMError: Check failed: it != type_definitions.end(): There is no definition of static_tensor_float32_*

[GitBox <gi...@apache.org>]

nolanliou edited a comment on issue #6268:
URL: https://github.com/apache/incubator-tvm/issues/6268#issuecomment-685459847


   @masahi Sample code to reproduce the problem of inputs' type not matching.
   ```
   import numpy as np
   import torch
   import tvm
   from tvm import relay
   import tvm.contrib.graph_runtime as runtime
   
   def fn(x):
       y = (x - 1).long()
       y = y.to(dtype=torch.float32)
       y = y[:, 0]
       return y
   
   jit = torch.jit.trace(fn, (torch.ones(4, 5, dtype=torch.int64),))
   torch.jit.save(jit, "test.jit")
   
   jit = torch.jit.load("test.jit", map_location=torch.device('cpu'))
   jit.eval()
   
   inputs = torch.randint(low=0, high=10, size=(4, 5), dtype=torch.int64)
   print(jit(inputs))
   
   
   shape_list = [('input', [4, 5])]
   
   mod, params = relay.frontend.pytorch.from_pytorch(jit,
                                                     shape_list,
                                                     default_dtype="float32")
   
   with tvm.transform.PassContext(opt_level=3):
       json, lib, params = relay.build(mod, target="llvm", params=params)
   
   ctx = tvm.cpu(0)
   inputs = tvm.nd.array(inputs.cpu().numpy().astype(np.int64), ctx)
   runtime = tvm.contrib.graph_runtime.create(json, lib, ctx)
   runtime.set_input(**params)
   runtime.set_input('input', inputs)
   runtime.run()
   tvm_result = runtime.get_output(0).asnumpy()
   print(tvm_result)
   ```
   
   There is no type of information in the JIT file, seems TVM cannot support the situation (load from JIT file) yet.


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org